Re: [PATCH v4 1/2] y2038: linux: Provide __timerfd_gettime64 implementation

[Adhemerval Zanella <adhemerval.zanella@linaro.org>]

On 07/01/2020 06:27, Lukasz Majewski wrote:

>> As a side note, now that arch-syscall patch is upstream should we
>> assume that for !__ASSUME_TIME64_SYSCALLS the __NR_timerfd_gettime64
>> should be defined (meaning that Linux supports time64 for all 32-bit
>> architectures)?
> 
> Only Linux version >= 5.1 supports 64 bit time on archs with __WORDSIZE
> = 32. I do guess (but I may be wrong here) that the arch-syscall is
> supposed to reflect the exact syscalls provided by kernel headers used
> for building (to help with validation of Y2038 patches).

The arch-syscall is now autogenerated from the latest kernel release
defined in build-many-glibcs.py. So the question is whether Linux
support and enforces time64 support on all and future 32-bit 
architectures or if there is still some missing ones (as it has
happen on some syscall additions, where some architecture lag
behind some releases).


> 
>>
>>> +  struct itimerspec its32;
>>> +  int retval = INLINE_SYSCALL_CALL (timerfd_gettime, fd, &its32);
>>> +  if (retval == 0)
>>> +    {
>>> +      value->it_interval = valid_timespec_to_timespec64
>>> (its32.it_interval);
>>> +      value->it_value = valid_timespec_to_timespec64
>>> (its32.it_value);
>>> +    }
>>> +
>>> +  return retval;
>>> +#endif
>>> +}  
>>
>>
>> Ok.
>>
>>> +
>>> +#if __TIMESIZE != 64
>>> +libc_hidden_def (__timerfd_gettime64)  
>>
>> Ok.
>>
>> As a side note, we should fix it on clock_{get,set}time to add the
>> missing libc_hidden_def.
> 
> The clock_gettime already has libc_hidden_def. The difference is that we
> use some compatibility code (after moving clock_gettime from librt to
> libc) instead of strong_alias (as it mimics the behavior from auto
> generated syscall wrapper).

I meant for the new time64 symbols.  Currently it is not an issue because
the internal time64 symbol is not exported and static linker uses the
internal __GI_ name for the symbol.  For instance, objdump -t on
clock_gettime.os on a 32-bit architecture (powerpc in this case) shows:

 00000144 g     F .text  00000088 __clock_gettime
 00000144 g     F .text  00000088 __clock_gettime_2
 00000000 g     F .text  00000144 .hidden __GI___clock_gettime64
 00000144 g     F .text  00000088 .hidden __GI___clock_gettime
 00000144 g     F .text  00000088 clock_gettime@@GLIBC_2.17
 00000144 g     F .text  00000088 clock_gettime@GLIBC_2.2

Where with a libc_hidden_def (__clock_gettime64) it shows:

 00000144 g     F .text  00000088 __clock_gettime
 00000144 g     F .text  00000088 __clock_gettime_2
 00000000 g     F .text  00000144 .hidden __GI___clock_gettime64
*00000000 g     F .text  00000144 __clock_gettime64
 00000144 g     F .text  00000088 .hidden __GI___clock_gettime
 00000144 g     F .text  00000088 clock_gettime@@GLIBC_2.17
 00000144 g     F .text  00000088 clock_gettime@GLIBC_2.2

The requirement of libc_hidden_def will de defined in the end if glibc
exports or not __clock_gettime64 on some header redirection or if 
clock_gettime64 would be suffice (with a {weak,strong}_alias).

However I do think we should fix it to avoid such confusion why there 
is a hidden_proto and not a hidden_def.